Robotic systems for retrieval of contaminated material from hazardous zones

韩国和英国的核工业都面临着相同的挑战，即访问大型放化工艺环境以执行远程干预任务。特别是，英国和韩国都确定了对无人系统的巨大需求，这些系统可以在危险到不可能有人进入的区域处理和回收受污染的材料。该项目将通过开发新型机器人系统来直接满足这一需求，该系统可以进入、监测和执行对核退役环境中的各种物体和材料的操纵操作，否则这些物体和材料将仍然无法进入和管理。UOB-Kairi-NL财团将为移动机器人机械手开发硬件、软件、算法和控制方法，包括配备手臂和末端执行器(可能包括手和/或切割设备)的无人驾驶车辆，这些机器人可以进入危险环境，在这些环境中对材料执行各种各样的操纵任务，并以受控的方式从环境中取回物体。此外，我们还将开发一种更小的儿童爬管机器人，它可以骑在母车上，并通过母车的操纵臂部署到管道上(在许多核设施中很常见)。儿童机器人的目的是检查在高度复杂和3D核电站环境中原本无法进入的区域，例如到达高、窄或杂乱的地方。此外，安装在儿童机器人上的摄像头可以提供有用的母车替代视图，促进对操作臂的自主“视觉伺服”控制，和/或由专业的人类操作员进行更好的遥操作控制。控制方法将是“半自主”、“远程自主”或“可变自主”之一，因此将超越以前在核环境中尝试的方法。传统上，安全关键型行业在允许将控制权从操作员移交给自主机器方面一直非常保守，转而依赖于直接的远程操作(例如，人类通过开关或喜悦棒控制机械臂的每个关节)。然而，越来越清楚的是，i)退役任务的巨大规模，以及ii)执行退役所需的机器人的复杂性和高自由度，意味着将需要某些类型的自主控制作为“操作员辅助”技术。例如，人类操作员应该能够用鼠标点击一个物体，并让机器人自主地抓住它，而不是人类试图直接控制两个或更多移动的基础马达、六个或更多的手臂马达和两个或更多的抓手手指。此外，自主感知方法，如通过计算视觉重建环境，将是远程人类操作员的态势感知以及在机器人上运行的自动规划和控制算法所必需的。我们将开发先进的计算机视觉和路径规划算法，实现机器人车辆的无碰撞导航，以及成功的自主手臂和手轨迹，以实现对任意形状的物体和材料的强健抓取。此外，我们将开发先进的动力学模型和控制方法，以促进高度动态的机器人动作，如爬升子机器人的支撑或摆动，或母亲移动机械手与其环境接触的强力动作，如切割和研磨物体，或拖动抓取的物体。总体目标是能够安全、无人驾驶地从危险区域回收受污染的材料。

Both Korean and UK nuclear industries share the same challenge of accessing large radio-chemical process environments to perform remote intervention tasks. In particular, both UK and Korea have identified a significant need for unmanned systems which can handle and retrieve contaminated materials in zones which are too hazardous to risk manned entries.This project will directly address this need by developing novel robotic systems which can enter, monitor, and carry out manipulative actions on a wide variety of objects and materials in nuclear decommissioning environments, which would otherwise remain inaccessible and unmanageable.The UoB-KAERI-NNL consortium will develop hardware, software, algorithms and control methods for a mobile robot manipulator, comprising an unmanned vehicle equipped with an arm and end-effectors (which could include hands and-or cutting devices), which can enter hazardous environments, perform a wide variety of manipulation tasks on materials inside those environments, and retrieve objects from the environment in a controlled fashion.Additionally, we will develop a smaller "child" pipe-climbing robot, which can ride on the mother vehicle and be deployed onto pipe-work (prevalent in many nuclear installations) via the mother vehicle's manipulator arm. The purpose of the child robot is to inspect zones which would otherwise be inaccessible in highly complex and 3D nuclear plant environments, for example to reach places that are high, narrow or cluttered. Additionally, cameras mounted on the child robot can provide useful alternative views of the mother-vehicle, facilitating autonomous "visual-servoing" control of the manipulator arm, and/or better tele-operative control by an expert human operator.The control approach will be one of "semi-autonomy", "tele-autonomy" or "variable-autonomy" which would therefore go beyond what has previously been attempted in nuclear environments. Traditionally, safety-critical industries have been very conservative about allowing the devolution of control from human operator to an autonomous machine, and have instead relied on direct tele-operation (e.g. a human controlling each joint of a robot arm by means of switches or joy-sticks). However, it is becoming increasingly clear that the combination of i) the vast scale of the decommissioning task, and ii) the complexity and high degrees-of-freedom of the robots needed to perform decommissioning, means that certain kinds of autonomous control will be required as "operator-assistance" technologies. For example, a human operator should be able to mouse-click on an object, and have the robot autonomously grasp it, rather than the human attempting to control two or more mobile base motors, six or more arm motors and two or more gripper fingers directly.Additionally, autonomous sensing approaches, such as 3D reconstruction of environments by computational vision, will be necessary for both situational awareness of the remote human operator, and automatic planning and control algorithms running on the robot. We will develop advanced computer vision and path-planning algorithms, which will enable collision-free navigation of the robot vehicle, and successful autonomous arm and hand trajectories to effect robust grasps on arbitrarily shaped objects and materials.Furthermore, we will develop advanced dynamics models and control methods to facilitate highly dynamic robot actions, such as braciation or swinging of the climbing child robot, or forceful actions of the mother mobile-maniplator with respect to contacts with its environment, for example cutting and grinding of objects, or dragging of grasped objects.The overall aim is to enable the safe, unmanned retrieval of contaminated materials from hazardous zones.